Real-time estimation of magnitude-square coherence for use in implantable devices

Magnitude square coherence (MSC) has been shown to be a useful technique in discriminating fibrillatory rhythms (FIB) from nonfibrillatory rhythms (NON-FIB). The authors present a method for estimating MSC in real-time for intracardiac electrograms using an overlapped moving discrete Fourier transform approach. Real-time discrimination of FIB and NONFIB is performed by using 32-point data segments and a Hanning window. Fibrillation is detected by using a simple threshold technique for mean MSC in the 0-60-Hz band. This method of real-time mean MSC estimation reliably discriminates FIB from NONFIB rhythms in a time period appropriate for use in implantable devices. A moving mean MSC offers a new perspective on cardiac dynamics.<>